Tool use and agents · Alibaba Qwen

Qwen3 4B Instruct 2507

Hugging Face: Qwen/Qwen3-4B-Instruct-2507 · checked 2026-10-13

Non-thinking 4B update with 256K context; a capable small agent.

Datasheet

Maker	Alibaba Qwen
Parameters	4.02 BVerifiedsafetensors.total from the HF API Source
Licence	Apache-2.0Verified
Access	Public download
Task (HF pipeline)	text-generation
First published on HF	2025-08-05
Last updated on HF	2025-09-17
Weights in memory	8.0 GB at 16-bit · 2.4 GB at ~4.85 bitsEstimateParameters × bits ÷ 8. Excludes context cache and runtime overhead; not every model has a 4-bit build.

Benchmark

35.68%

BFCL V4, overall accuracy · #3 of 3 scored in this category

Qwen3-4B-Instruct-2507 (FC) row, rank 54. Source

Other published results

BFCL-v3 (vendor-reported): 61.9%Sourced Source

See the full tool use and agents ranking

Used in

Answering the phone for a small business

The description summarises the model card at https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507. Parameters and licence are read from the Hugging Face API; the weights-in-memory line is our arithmetic, not a measurement.